Evaluating the Performance of Algorithms in Estimating the Chl-a Concentration of Lake Bafa

Abstract

Monitoring and estimating pigment concentrations in water bodies have a critical role in early intervention or investigation of causes for prevention. Remote sensing data are the most effective alternative due to its advantages as effortless, requiring less labor, and displaying large areas in a single frame. Analyzing and estimating Chlorophyll-a (Chl-a) concentrations constitute the most important research topics in water bodies because all phytoplankton contain Chl-a. In this study, we evaluated the performance of algorithms in estimating the Chl-a concentration of Lake Bafa based on Sentinel 2 bands which are simulated from in-situ reflectance data. We used 1/R665xR705, 1/R665-1/R705, (1/R665-1/R705) x R740, R705/(R560+R665), so called M09, G09-2B, G09-3B, K07, respectively and Normalized Difference Chlorophyll Index (NDCI) algorithms for evaluation. Water samples and in-situ measurements were collected and obtained in two field campaigns. Bands of Sentinel 2 were then simulated from in-situ reflectance data and used to calibrate and validate models for Chl-a estimation. R² values of 0.679, 0.749, 0.395, 0.726, and RMSE values of 0.7 and 1.882, 1.663, 1.737, and 1.818 μg/L have been obtained for M09, G09-2B, G09-3B, K07, and NDCI algorithms, respectively. Sentinel 2 images have been used for map validation. Our results show that M09 and NDCI algorithms performed better in estimating Chl-a compared to the other three algorithms for our data range at Lake Bafa.